🤖 AI Summary
This work addresses visual prompt-based keyword localization (VPKL) for real-world low-resource languages—such as Yoruba—where no textual transcriptions of speech are available. Method: We propose the first transcription-free VPKL framework, leveraging vision-guided speech keyword localization without any speech-text alignments. It introduces a transcription-free few-shot positive/negative sample mining mechanism, integrated with vision–speech contrastive learning and cross-modal alignment modeling to mitigate severe annotation scarcity. Results: Experiments show only marginal performance degradation on English, confirming robustness; more importantly, we achieve the first feasible VPKL results on Yoruba—substantially outperforming baselines—and demonstrate strong effectiveness and generalizability under authentic low-resource conditions. The core contribution is the construction of the first transcription-free VPKL framework, establishing a novel paradigm for low-resource multimodal speech understanding.
📝 Abstract
Given an image query, visually prompted keyword localisation (VPKL) aims to find occurrences of the depicted word in a speech collection. This can be useful when transcriptions are not available for a low-resource language (e.g. if it is unwritten). Previous work showed that VPKL can be performed with a visually grounded speech model trained on paired images and unlabelled speech. But all experiments were done on English. Moreover, transcriptions were used to get positive and negative pairs for the contrastive loss. This paper introduces a few-shot learning scheme to mine pairs automatically without transcriptions. On English, this results in only a small drop in performance. We also - for the first time - consider VPKL on a real low-resource language, Yoruba. While scores are reasonable, here we see a bigger drop in performance compared to using ground truth pairs because the mining is less accurate in Yoruba.